Work

Real systems, doing real work

Client names stay confidential — that's a promise we'd make to you too. Here's what was built, why, and how it works.

Document pipelineConstruction trades

Spec reports into finished job sheets

Problem
A specialty trade contractor receives long project specification reports from its customers. Each report covers every trade on the job, the format changes from customer to customer, and staff were reading them line by line so nothing in their scope got missed.
Built
A rule-based pipeline that reads the report, pulls out only the items in that trade's scope, applies the company's own conventions — how products are named, what gets included, how headers and job numbers are formatted — and produces a finished, print-ready job sheet.
Why rules
General-purpose AI tools given the same documents were less reliable: they skipped items or invented detail. Encoding the team's know-how as explicit rules made the output consistent and checkable.
Result
Line-by-line reading replaced by an automated first pass that a person reviews. Same format every time, and the company's tribal knowledge now lives in a system instead of in one person's head.
Data privacyAI prep

A scrubber that runs before the AI does

Problem
Purchase orders downloaded from a supplier portal needed to be turned into internal orders, and AI could help — but the files carried pricing, contact details and session links that shouldn't be pasted into an AI tool.
Built
A small tool that strips sensitive fields from the files first, so only the information needed to build the order ever reaches the AI step.
Result
Faster order entry without leaking business data. A public version of the idea is our free data cleaner — and the full version comes as a bonus with our document projects.
InfrastructureMonitoring

A self-hosted automation server

Problem
Hosted automation tools charge per task and keep your data on someone else's platform. Running your own server is cheaper at volume — but only if it's monitored, backed up and secure.
Built
One server running the n8n workflow engine, Typebot chatbots, uptime monitoring, cookie-free analytics, website change tracking and PDF tools, all behind HTTPS. Every failed workflow sends an alert to a phone with the error and a link. Workflows are backed up off the server nightly, and a morning report summarizes failures, backups and anything that changed.
Result
Unlimited workflows for a flat monthly server cost, problems surfaced within seconds, and a documented recovery path. It's also what this website runs on.
Content opsGoogle Drive

One image in, every platform out

Problem
Every social platform wants a different image size. Posts were failing or getting cropped badly because one image was being pushed everywhere.
Built
Drop an image into a Drive folder. Within about a minute it's resized for TikTok, Instagram, Facebook and Pinterest — artwork fitted without cropping on a blurred background — saved into per-platform folders, and the original is filed as processed.
Result
No more rejected or awkwardly cropped posts, and a scheduling tool picks the right size from the right folder automatically.
MonitoringSlack

Competitor and release watch

Problem
Keeping an eye on competitors' product pages and on software updates meant remembering to check a dozen websites.
Built
Change tracking on chosen pages, with a short plain-text summary of what changed posted to a team channel — added lines, removed lines, nothing else.
Result
Price and catalog moves show up on their own, and security updates for the tools in use are never missed.

Got a process like one of these?

Tell us about it and get a fixed quote.

Start a conversation